Summary of Graph Transductive Defense: a Two-stage Defense For Graph Membership Inference Attacks, by Peizhi Niu et al.
Graph Transductive Defense: a Two-Stage Defense for Graph Membership Inference Attacks
by Peizhi Niu, Chao Pan, Siheng Chen, Olgica Milenkovic
First submitted to arxiv on: 12 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary In this paper, researchers address the vulnerability of graph neural networks (GNNs) to membership inference attacks (MIA) in graph transductive learning settings. GNNs are widely used for tasks like social network analysis and medical data processing, but MIA can compromise their privacy by identifying whether a record was part of the model’s training data. The authors propose Graph Transductive Defense (GTD), a two-stage defense mechanism that combines a train-test alternate training schedule and flattening strategy to reduce the difference between training and testing loss distributions. Experimental results show that GTD outperforms LBP, with a 9.42% decrease in attack AUROC and an 18.08% increase in utility performance on average. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary GNNs are special types of artificial intelligence used for analyzing data that is connected like social networks or medical records. These GNNs can be tricked into revealing if they have seen certain information before, which is bad because it can compromise people’s privacy. The researchers who wrote this paper wanted to fix this problem by creating a new way to defend against these attacks in situations where the GNN is learning from data that is not just for one specific group. They came up with an idea called Graph Transductive Defense (GTD) that combines two steps: training and testing, and then making sure everything looks similar. It worked really well and can be used with other classification models without causing too much extra work. |
Keywords
» Artificial intelligence » Classification » Gnn » Inference
